1 /*
2 * CDDL HEADER START
3 *
4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License, Version 1.0 only
6 * (the "License"). You may not use this file except in compliance
7 * with the License.
8 *
9 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
10 * or http://www.opensolaris.org/os/licensing.
11 * See the License for the specific language governing permissions
12 * and limitations under the License.
13 *
14 * When distributing Covered Code, include this CDDL HEADER in each
15 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
16 * If applicable, add the following below this CDDL HEADER, with the
17 * fields enclosed by brackets "[]" replaced with your own identifying
18 * information: Portions Copyright [yyyy] [name of copyright owner]
19 *
20 * CDDL HEADER END
21 */
22 /*
23 * Copyright 2002 Sun Microsystems, Inc. All rights reserved.
24 * Use is subject to license terms.
25 */
26
27 #pragma ident "%Z%%M% %I% %E% SMI"
28
29 #include <sys/types.h>
30 #include <sys/kmem.h>
31 #include <sys/conf.h>
32 #include <sys/sysmacros.h>
33 #include <netinet/in.h>
34 #include <netinet/in_systm.h>
35 #include <netinet/ip6.h>
36 #include <inet/common.h>
37 #include <inet/ip.h>
38 #include <inet/ip6.h>
39 #include <ipp/meters/meter_impl.h>
40
41 /*
42 * Module : Single or Two Rate Metering module - tokenmt
43 * Description
44 * This module implements the metering part of RFC 2698 & 2697. It accepts the
45 * committed rate, peak rate (optional), committed burst and peak burst for a
46 * flow and determines if the flow is within the cfgd. rates and assigns
47 * next action appropriately..
48 * If the peak rate is provided this acts as a two rate meter (RFC 2698), else
49 * a single rate meter (RFC 2697). If this is a two rate meter, then
50 * the outcome is either green, red or yellow. Else if this a single rate
51 * meter and the peak burst size is not provided, the outcome is either
52 * green or red.
53 * Internally, it maintains 2 token buckets, Tc & Tp, each filled with
54 * tokens equal to committed burst & peak burst respectively initially.
55 * When a packet arrives, tokens in Tc or Tp are updated at the committed
56 * or the peak rate up to a maximum of the committed or peak burst size.
57 * If there are enough tokens in Tc, the packet is Green, else if there are
58 * enough tokens in Tp, the packet is Yellow, else the packet is Red. In case
59 * of Green and Yellow packets, Tc and/or Tp is updated accordingly.
60 */
61
62 int tokenmt_debug = 0;
63
64 /* Updating tokens */
65 static void tokenmt_update_tokens(tokenmt_data_t *, hrtime_t);
66
67 /*
68 * Given a packet and the tokenmt_data it belongs to, this routine meters the
69 * ToS or DSCP for IPv4 and IPv6 resp. with the values configured for
70 * the tokenmt_data.
71 */
72 int
tokenmt_process(mblk_t ** mpp,tokenmt_data_t * tokenmt_data,ipp_action_id_t * next_action)73 tokenmt_process(mblk_t **mpp, tokenmt_data_t *tokenmt_data,
74 ipp_action_id_t *next_action)
75 {
76 uint8_t dscp;
77 ipha_t *ipha;
78 ip6_t *ip6_hdr;
79 uint32_t pkt_len;
80 mblk_t *mp = *mpp;
81 hrtime_t now;
82 enum meter_colour colour;
83 tokenmt_cfg_t *cfg_parms = tokenmt_data->cfg_parms;
84
85 if (mp == NULL) {
86 tokenmt0dbg(("tokenmt_process: null mp!\n"));
87 atomic_add_64(&tokenmt_data->epackets, 1);
88 return (EINVAL);
89 }
90
91 if (mp->b_datap->db_type != M_DATA) {
92 if ((mp->b_cont != NULL) &&
93 (mp->b_cont->b_datap->db_type == M_DATA)) {
94 mp = mp->b_cont;
95 } else {
96 tokenmt0dbg(("tokenmt_process: no data\n"));
97 atomic_add_64(&tokenmt_data->epackets, 1);
98 return (EINVAL);
99 }
100 }
101
102 /* Figure out the ToS/Traffic Class and length from the message */
103 if ((mp->b_wptr - mp->b_rptr) < IP_SIMPLE_HDR_LENGTH) {
104 if (!pullupmsg(mp, IP_SIMPLE_HDR_LENGTH)) {
105 tokenmt0dbg(("tokenmt_process: pullup error\n"));
106 atomic_add_64(&tokenmt_data->epackets, 1);
107 return (EINVAL);
108 }
109 }
110 ipha = (ipha_t *)mp->b_rptr;
111 if (IPH_HDR_VERSION(ipha) == IPV4_VERSION) {
112 /* discard last 2 unused bits */
113 dscp = ipha->ipha_type_of_service;
114 pkt_len = ntohs(ipha->ipha_length);
115 } else {
116 ip6_hdr = (ip6_t *)mp->b_rptr;
117 /* discard ECN bits */
118 dscp = __IPV6_TCLASS_FROM_FLOW(ip6_hdr->ip6_vcf);
119 pkt_len = ntohs(ip6_hdr->ip6_plen) +
120 ip_hdr_length_v6(mp, ip6_hdr);
121 }
122
123 /* Convert into bits */
124 pkt_len <<= 3;
125
126 now = gethrtime();
127
128 mutex_enter(&tokenmt_data->tokenmt_lock);
129 /* Update the token counts */
130 tokenmt_update_tokens(tokenmt_data, now);
131
132 /*
133 * Figure out the drop preced. for the pkt. Need to be careful here
134 * because if the mode is set to COLOUR_AWARE, then the dscp value
135 * is used regardless of whether it was explicitly set or not.
136 * If the value is defaulted to 000 (drop precd.) then the pkt
137 * will always be coloured RED.
138 */
139 if (cfg_parms->tokenmt_type == SRTCL_TOKENMT) {
140 if (!cfg_parms->colour_aware) {
141 if (pkt_len <= tokenmt_data->committed_tokens) {
142 tokenmt_data->committed_tokens -= pkt_len;
143 *next_action = cfg_parms->green_action;
144 } else if (pkt_len <= tokenmt_data->peak_tokens) {
145 /*
146 * Can't do this if yellow_action is not
147 * configured.
148 */
149 ASSERT(cfg_parms->yellow_action !=
150 TOKENMT_NO_ACTION);
151 tokenmt_data->peak_tokens -= pkt_len;
152 *next_action = cfg_parms->yellow_action;
153 } else {
154 *next_action = cfg_parms->red_action;
155 }
156 } else {
157 colour = cfg_parms->dscp_to_colour[dscp >> 2];
158 if ((colour == TOKENMT_GREEN) &&
159 (pkt_len <= tokenmt_data->committed_tokens)) {
160 tokenmt_data->committed_tokens -= pkt_len;
161 *next_action = cfg_parms->green_action;
162 } else if (((colour == TOKENMT_GREEN) ||
163 (colour == TOKENMT_YELLOW)) &&
164 (pkt_len <= tokenmt_data->peak_tokens)) {
165 /*
166 * Can't do this if yellow_action is not
167 * configured.
168 */
169 ASSERT(cfg_parms->yellow_action !=
170 TOKENMT_NO_ACTION);
171 tokenmt_data->peak_tokens -= pkt_len;
172 *next_action = cfg_parms->yellow_action;
173 } else {
174 *next_action = cfg_parms->red_action;
175 }
176 }
177 } else {
178 if (!cfg_parms->colour_aware) {
179 if (pkt_len > tokenmt_data->peak_tokens) {
180 *next_action = cfg_parms->red_action;
181 } else if (pkt_len > tokenmt_data->committed_tokens) {
182 /*
183 * Can't do this if yellow_action is not
184 * configured.
185 */
186 ASSERT(cfg_parms->yellow_action !=
187 TOKENMT_NO_ACTION);
188 tokenmt_data->peak_tokens -= pkt_len;
189 *next_action = cfg_parms->yellow_action;
190 } else {
191 tokenmt_data->committed_tokens -= pkt_len;
192 tokenmt_data->peak_tokens -= pkt_len;
193 *next_action = cfg_parms->green_action;
194 }
195 } else {
196 colour = cfg_parms->dscp_to_colour[dscp >> 2];
197 if ((colour == TOKENMT_RED) ||
198 (pkt_len > tokenmt_data->peak_tokens)) {
199 *next_action = cfg_parms->red_action;
200 } else if ((colour == TOKENMT_YELLOW) ||
201 (pkt_len > tokenmt_data->committed_tokens)) {
202 /*
203 * Can't do this if yellow_action is not
204 * configured.
205 */
206 ASSERT(cfg_parms->yellow_action !=
207 TOKENMT_NO_ACTION);
208 tokenmt_data->peak_tokens -= pkt_len;
209 *next_action = cfg_parms->yellow_action;
210 } else {
211 tokenmt_data->committed_tokens -= pkt_len;
212 tokenmt_data->peak_tokens -= pkt_len;
213 *next_action = cfg_parms->green_action;
214 }
215 }
216 }
217 mutex_exit(&tokenmt_data->tokenmt_lock);
218
219 /* Update Stats */
220 if (*next_action == cfg_parms->green_action) {
221 atomic_add_64(&tokenmt_data->green_packets, 1);
222 atomic_add_64(&tokenmt_data->green_bits, pkt_len);
223 } else if (*next_action == cfg_parms->yellow_action) {
224 atomic_add_64(&tokenmt_data->yellow_packets, 1);
225 atomic_add_64(&tokenmt_data->yellow_bits, pkt_len);
226 } else {
227 ASSERT(*next_action == cfg_parms->red_action);
228 atomic_add_64(&tokenmt_data->red_packets, 1);
229 atomic_add_64(&tokenmt_data->red_bits, pkt_len);
230 }
231
232 return (0);
233 }
234
235 void
tokenmt_update_tokens(tokenmt_data_t * tokenmt_data,hrtime_t now)236 tokenmt_update_tokens(tokenmt_data_t *tokenmt_data, hrtime_t now)
237 {
238 tokenmt_cfg_t *cfg_parms = (tokenmt_cfg_t *)tokenmt_data->cfg_parms;
239 hrtime_t diff = now - tokenmt_data->last_seen;
240 uint64_t tokens;
241
242 switch (cfg_parms->tokenmt_type) {
243 case SRTCL_TOKENMT:
244 tokens = (cfg_parms->committed_rate * diff) /
245 METER_SEC_TO_NSEC;
246
247 /*
248 * Add tokens at the committed rate to
249 * committed_tokens. If they are in excess of
250 * the committed burst, add the excess to
251 * peak_tokens, capped to peak_burst.
252 */
253 if ((tokenmt_data->committed_tokens + tokens) >
254 cfg_parms->committed_burst) {
255 tokens = tokenmt_data->committed_tokens
256 + tokens -
257 cfg_parms->committed_burst;
258 tokenmt_data->committed_tokens =
259 cfg_parms->committed_burst;
260 tokenmt_data->peak_tokens =
261 MIN(cfg_parms->peak_burst,
262 tokenmt_data->peak_tokens +
263 tokens);
264 } else {
265 tokenmt_data->committed_tokens +=
266 tokens;
267 }
268 break;
269 case TRTCL_TOKENMT:
270 /* Fill at the committed rate */
271 tokens = (diff * cfg_parms->committed_rate) /
272 METER_SEC_TO_NSEC;
273 tokenmt_data->committed_tokens =
274 MIN(cfg_parms->committed_burst,
275 tokenmt_data->committed_tokens + tokens);
276
277 /* Fill at the peak rate */
278 tokens = (diff * cfg_parms->peak_rate) /
279 METER_SEC_TO_NSEC;
280 tokenmt_data->peak_tokens =
281 MIN(cfg_parms->peak_burst,
282 tokenmt_data->peak_tokens + tokens);
283 break;
284 }
285 tokenmt_data->last_seen = now;
286 }
287